Abstract
The purpose of this paper is to present a signal detection approach where the signal to be detected is corrupted by a colored random noise whose statistics are unknown.
First, the problem of identifying the unknown noise statistics is solved by computing the likelihood-ratio functions under the a priori assumption that the unknown noise covariance has one of the preassigned forms. Secondly, the detection procedure is given based on the identification of noise statistics by using the likelihood-ratio concept. The latter half of this paper is devoted to present an application of the general theory to practical problems in the case where preassigned signals are contaminated by a real observed random noise with unknown noise statistics. Results of computer experiments are shown to demonstrate the feasibility of the proposed method.
